European Journal of Biomedical and Pharmaceutical Sciences

BRAIN ABNORMALITY CLASSIFICATION AND DETECTION BY PARTICLE SWARM OPTIMIZATION-NEURAL NETWORK IN MR BRAIN IMAGES

Shrikant Burje*, Prof. Dr. Sourabh Rungta and Prof. Dr. Anupam Shukla

ABSTRACT

Knowledge-based medical intelligence system played a key role in the diagnosis of diseases represents crucial role in the design of expert systems in medical diagnosis. The main intention of this proposal is to detect and classification of brain diseases from the Magnetic resonance image. Magnetic Resonance medical image analysis is most important tasks in clinical diagnosis. The abnormalities in brain define by tumor benign and malignant. In the analysis of brain MR images, we review the best methods used for feature extraction and reduction. The proposed method integration of Discrete Wavelet Transform, Principal Component Analysis (PCA), Feed forward neural network Classifier (FFNC) and PSO. The PSO algorithm is contributing towards minimizing the objective. The methodology embraces various mechanisms like features extraction, image segmentation, feature selection and dimension reduction. We considered approaches to design the expert system for diagnosis of abnormalities in the brain. The various basic machine learning algorithm very often used in the artificial neural network. In Proposed hybrid method, feed-forward neural networks based on particle swarm optimization achieve the highest classification accuracy 99.33%. This study brings out the performance analysis of the hybrid system regarding accuracy and computational cost.

Keywords: PSO, MR, PCA, FFNC, DWT.